In general, a digital communication transmitter includes a source for information, which could be e.g. an MPEG audio encoder for digital broadcasting and/or an MPEG video encoder for digital television. Typically, the output data of the communication source, which are present in form of a digital bit stream, are then encoded using a channel encoder to add redundancy to the bit stream, which serves to help overcome transmission errors in the receiver. Subsequently, the channel-encoded digital bit stream is fed into a so-called “interleaver” which changes the order of data according to an algorithm which is known to the receiver to prevent so-called burst errors in the transmission channel from resulting in a loss of a greater, coherent part of the communication, but only in smaller and short losses which are distributed over a longer period of time. Using a so-called mapper, and depending on the type of modulation, the nested bit stream at the output of the interleaver will now be mapped to modulation symbols.
If no modulation is used and if the digital bit stream is, so to speak, sent directly without any modulation, the mapper as well as the subsequently used modulator will be omitted.
If, however, use is made of a modulation method, e.g. a multi-carrier modulation process, the mapper is followed by a modulator modulating the modulation symbols onto the communication carriers.
Recently, the OFDM process for digital broadcasting applications has become ever more popular. In this process a multiplicity of sub-carriers is used onto which the modulation symbols produced by the mapper are modulated. In this case, the modulation process is an inverse discrete Fourier transform to generate, as is generally known, a discrete time signal from the many modulated carriers. The discrete, usually complex time signal occurs in the form of complex sample values or samples, which are then fed into an interpolation low pass in order to remove the periodically repeating spectral components. A signal is applied across the output of the interpolation low pass, which is typically modulated onto an RF carrier frequency by way of a complex IQ modulator to receive a signal which will finally be fed into a transmitting amplifier which will supply the amplified signal to an antenna by which the signal will finally be emitted.
In broadcasting applications, due to the high power requirements, valve amplifiers, such as klystrons or travelling wave tubes, are typically used at the output of the amplifier. Should smaller powers suffice, such as in mobile radio, where there exists a dense network of transmitters, transistor amplifiers may also be used.
Both transistor amplifiers and valve amplifiers have in common that they are linear only in a certain input power range and, regarding increasing input powers, have a decreasing output power curve which finally exhibits a constant behaviour when the amplifier has been completely saturated. In other words, the amplifier exerts non-linear distortions on the input signal owing to its non-linear characteristic curve towards higher input powers.
In situations where a certain frequency band has been assigned for a certain transmitting application by e.g. a state licensing authority, regulations exist in that the transmit signal for the specific licensed transmitting application may only have power in a prescribed band, but, however, may have no or only small power outside the assigned band. Power outside the assigned band is also referred to as out-of-band radiation.
As has already been mentioned, in the case of higher input powers, the non-linear characteristic curve of the amplifier leads to non-linear distortions, such, that higher harmonic waves are generated by the amplifier which are no longer within the assigned band but which reside outside of the band and which may be measured as out-of-band radiation.
It is known that these non-linear distortions result in a relatively white spectrum. If the input signal into the amplifier is still band-limited—a situation which can be assumed in the case of OFDM modulation—the output signal then has power outside of this band.
To avoid this, i. e. to observe regulations by the licensor for tolerance schemes established for frequency bands, i. e. how much of the out-of-band radiation is still acceptable outside the assigned band, the input voltage into the amplifier should rarely or not at all exceed the maximum input voltage for a distortion-free amplification. In other words, this means that the maximum input voltage which is actually available should be as low as possible. If the maximum input voltage is always smaller than the maximum voltage, at which the amplifier is on the verge of operating in the non-linear range or amplifies only very little in the non-linear range with its out-of-band radiation ranging below the permitted value, distortions will never occur which will result in a higher than permitted out-of-band radiation.
Disadvantages of the above described OFDM process include the typically occurring great peak to average power ratio which is also referred to as PAR. In a graphic description, large peaks may occur in the time signal, i. e. in the OFDM symbol after the IDFT, when the carriers are being occupied so unfavourably, that, at a certain point in time, e.g. all of the 256 OFDM sub-carriers are superimposed in a constructive manner. In this case, a large signal peak will occur which may well be 10 to 20 dB above mean signal power. In order to keep to the permitted out-of-band radiation, a high power reserve is typically retained in the transmitting amplifier, which is also referred to as “power back-off”. In other words, the amplifier is operated at an operating point which is set so low that even a high power peak still lies in the linear range of the amplifier.
This mode of operation of an amplifier represents an extremely inefficient mode of operation in which the amplifier requires much supply power but provides only a relatively low output power. The demand for reduced out-of-band radiation in connection with high peak values in the time signal, which do not only occur in an OFDM modulation but also e.g. in a one-carrier process by an impulse former, i. e. which may generally occur during a filtering process, results in a need for more expensive amplifiers which have to be operated with a great amount of power reserve and which provide low efficiency. It is efficiency, however, which also represents an issue being paid ever more attention, especially for smaller battery-powered systems, particularly with respect to mobile radio and limited storage capacity accumulators used therein.
WO 98/10567 A1 relates to a method for reducing the peak value factor in digital transmission methods. Here, the basic idea is to take precautions in the digital domain to prevent high signal peaks from occurring in the time signal, such, that lower power reserves will suffice for the transmitting amplifier without any greater than permissible out-of-band radiation occurring. The known concept is generally referred to as “selected mapping”. Selected mapping or SLM actually means only that different possible signals, which may also be referred to as representative or candidate transmit sequences, are generated in any manner U from a message to be transmitted, i. e. an information word or, generally speaking, a vector of data bits. However, not all of these signals are transmitted. Instead, a special signal is selected as the transmit signal. In particular, each transmit signal has a peak value to be measured. The candidate transmit signal with the lowest peak value will finally be selected and transmitted as the actual transmit signal.
On the receiver side, the object is to find out (a) which message is available and (b) which of the U possible representatives was transmitted per message. The receiver has two possibilities to find this out. First, by means of page information which is reported somehow or other from the transmitter to the receiver and which relates to which of the U candidate transmit sequences was selected. What is disadvantageous about this method is the fact that this page information is transmitted explicitly and, in particular, that this page information has to be given special protection from transmission errors. In the case of digital broadcasting, where the channel may be disturbed in some way or other and such a disturbance is thus not easy to predict, this represents a critical issue.
If the page information is received incorrectly, error-free processing of the receiving signal is also no longer possible. This fact renders this way of signalising using page information relatively awkward.
Another possibility is to perform the concept without transmitting the page information. Solely by means of the receiving signal, the receiver then has to find out which message of M messages, i. e. which modulation symbol from a fixed number of modulation symbols, is available and which of the U candidate transmit sequences was transmitted per message. Instead of M possible transmit signals or modulation symbols from which the receiver has to choose the most likely signal or symbol, M×U possible transmit signals will now have to be considered, if no page information is used. This may cause the error probability to rise considerably. Furthermore, a suitable selection of U possible candidate transmit sequences per message represents a critical issue. So far, there has been a lack of a favourable approach for this problem, so that even the transmission without explicit page information has not yet proved itself suitable for practice.
A special implementation of SLM, which is also described in WO 98/10567 A1, is referred to as PTS concept (PTS=Partial Transmit Sequences). The U candidate transmit sequences are obtained the following way. The transmit signal is available, i. e. as a vector of complex-valued elements, before a final linear filtering process, e.g. a spectral shaping in one-carrier processes or an inverse Fourier transform in OFDM. The vector will now be partitioned into subsets, i. e. in partial transmit sequences. The elements of each subset may then be multiplied with the same complex number having the magnitude of 1. In the complex plane, these elements are all rotated about the same angle. Then, the final linear filtering operation will be effected through which the high peak values are typically generated first. Through the free choice of the complex number used for multiplication, it is now possible to generate the plurality of candidate transmit sequences. In this case, too, there are the two possibilities, as have been explained above, which include the option to operate with or without page information. Yet, the difference here is that it is possible to operate without transmitting explicit page information. This is possible since the information to be transmitted is not transmitted absolutely into the complex-valued elements of the time-discrete transmit signal, but in quotients of successive elements of the same subset. In general, this concept is referred to as differential precoding. Since all of the elements of one subset have been multiplied by the same complex number, the quotient of two successive elements of the same subset will remain the same for each U candidate transmit sequence. Accordingly, the receiver only has to calculate these quotients and is thus given back the transmitted information.
FIG. 10 shows a basic block diagram of one portion of a transmitter for transmit sequences generated according to a code multiplex process. An exemplary code multiplex process is known in the art as a CDMA process (CDMA=Code Division Multiple Access). CDMA is a long standing transmission technique wherein the information to be transmitted which is digitally available in the base band is spectrally “spread” during modulation, wherein this spreading is carried out using a unique code available for each individual connection, i.e. for each individual user or information channel. Thus, this technique is also called “Spread Spectrum”. The useful information is herefore weighted with a predetermined, typically higher-frequency bit sequence for every information channel, which is also referred to as code sequence. Thus, the occupied transmit bandwidth widens and/or the energy of this signal is distributed to a larger bandwidth.
This procedure is repeated for several information channels, and the thus obtained channel sequences are then added in order to form a transmit sequence. If a user knows the code for an information channel, then he can reconstruct this information channel from a received transmit sequence using a correlator which may be implemented in the form of a “matched filter”. The implementation of the matched filter hereby depends on the respective code sequence of the information channel.
One advantage of the CDMA technique generally is a continuously good quality and a higher capacity with a lower power consumption regarding the transmitter.
Due to these advantages of the CDMA process it is assumed, that the CDMA process will be used by the next mobile communication generation called UMTS, and which is the descendant of GSM. In particular, the CDMA process is used in the so-called downlink. Downlink hereby means, that the base stations and not the mobile parts are regarded as transmitters. The radio link from a base station to a mobile part is also referred to as downlink, while the radio link from a mobile part to a base station is referred to as uplink.
In particular, FIG. 10 shows a CDMA-stage. For reasons of clarity, in FIG. 10 only three information channels 102, 104 and 106 are shown. The three information channels may for example be three separate communication links between mobile radio users who all want to communicate via one radio channel. Information or messages in the respective information channels are optionally submitted to a usual channel encoding and a subsequent nesting, as it is shown by the dashed lines 108. This is illustrated in FIG. 10 by an encoder block 108a referred to as ENC. Typically the encoder block 108a will carry out a forward error correction, also referred to as FEC. An interleaver 108b is connected downstream to the encoder 108a designated with ILV, and which scrambles the bit sequence output by the encoder 108a in order to make the whole system insensitive against so-called burst errors.
The somehow encoded and nested information of the first information channel is hereupon fed into a means 110 for being weighted with a code sequence. An own code sequence is allocated to every information channel. This means, that a first code sequence 112 is associated with a first information channel 102, that a second code sequence 114 is associated with the second information channel 104 and a third code sequence 116 is associated with the third information channel 106. The three code sequences need at least to be different from each other, so that the information channels may be separated again. The best separation and therefore the best correlation peaks at the output of a correlator in a receiver are reached when the three code sequences are all orthogonal to each other, such that in a correlation with one code sequence only correlation peaks occur, if this code sequence is present in the examined signal in a way, that for the other code sequences which are also present in the examined signal no correlation peaks occur due to the orthogonality properties. As it is known, so-called pseudo noise sequences are used as code sequences which have the property that they comprise a relatively white spectrum and may on the other hand be produced by a feed-back shift register which is started depending on a certain output value.
At the output of the weighting means a so-called channel sequence is presented which generally speaking includes the information of the information channel weighted with the code sequence. The channel sequences of the individual sub-channels are then combined in one means 120, wherein this combination is typically carried out using a simple addition. On the output side a transmit sequence is applied then which is upconverted, amplified and emitted via an antenna or fed into a wire-bonded transmission channel by ways known in the art.
In the following the functioning of the weighting means 110 for weighting information using a code sequence is discussed. The information is for example applied at the output of the interleaver 108b or, however, if the same is not present, as information per se in the form of a sequence of individual bits. These bits may comprise a value of +1 for example for the information “1” and a value of −1 for the information “0”. A bit having a value of −1 leads to the fact that at the output of the weighting means 110 the code sequence itself is present, while a bit having a value of −1 causes that at the output of the weighting means 110 the inverted code sequence occurs, i.e., the original code sequence phase shifted by 180 degrees. When a sequence of bits is present at the input of the weighting means 110, there will be a sequence of positive or negative code sequences present at the output of the weighting means 110, which together form a channel sequence for this information channel. The weighting means 110 therefore causes that the channel sequence at the output of the weighting means 110 comprises a length of m×n bit, when the code sequence is n bit long and when a bit sequence at the output of the interleaver is m bit long.
The messages for the individual users are therefore optionally encoded and interleaved, then the weighting means 110 follows, which may generally also be referred to as mapper, as it generally carries out a mapping of the m bit to the m×n bit. As it is illustrated in FIG. 10, the code sequences differ for every information channel, which is critical for the CDMA process, so that the individual information channels may be separated again.
The different channel sequences of the different information channels are added in their combination means 120, as it was described above. This adding up of the individual channel sequences may cause the transmit sequence comprising high signal peaks. This problem becomes more critical especially when not only a small number of information channels is used, but when a large number of information channels needs to be added in order to obtain the transmit sequence. In the worst case all channel sequences have for example a positive value at a certain time, which results in a high positive peak value of the transmit sequence.
For this problem of high signal peaks in a CDMA transmit sequence there has only been the solution of operating the transmit amplifier with a large reserve, which leads to the mentioned disadvantages, as was discussed above.
A simple combination of the SLM-concept and/or the PTS concept as a sub-group of the SLM concept with the CDMA process, i.e. that a plurality of channel sequences is produced instead of one channel sequence, among which the one with the lowest peak value is selected then, leads to no drastic solution of the problem. The reason for this is, that the channel sequences are independent from each other for the individual information channels, such that it cannot be assumed that the most favourable candidate channel sequences from the individual sub channels will form a transmit sequence together which differentiates itself by a low peak value and/or by a low out-of-band radiation, i.e. which is the optimum.